DOC/BLD: fix annoying sphinx bugs

[cpcloud]

closes #3721. cc @jreback

[jreback]

this is in...thanks! built very nicey

[cpcloud]

crap...still a few more...that ok? can submit a separate pr, but i guess if they are building nicely for u that's good. i still get deprecation warnings about reset_printoptions which can easily be fixed

[jreback]

oh really....hmm..let me try to clean and build again (but I didn't see that before)....

why don't you put it in one that changes the warning on the install page to basically say install 4.02?

[cpcloud]

ok will do

[cpcloud]

running test suite now to make sure i didn't "overvim" everything by using a regex and :%s

[cpcloud]

oh darn it all to heck. now i get warnings about set_printoptions (easily fixed) and TimeSeries broadcasting. bugs better run and hide

[cpcloud]

screw it i am just running python make.py clean before the build

[jreback]

no...some of these are just warnings printed out by various whatsnews, so ok (on the Timeseries broadcasting)...maybe set_printoptions (as that is deprecated)

[cpcloud]

i'm getting ipython ext attrerrors now that is def a bug. of course the file name contains a hex value so that's useful for tracking it down

[cpcloud]

this doesn't work because x.item() is a long.

from datetime import datetime, timedelta
s  = Series(date_range('2012-1-1', periods=3, freq='D'))
td = Series([ timedelta(days=i) for i in range(3) ])
df = DataFrame(dict(A = s, B = td))
y = s - s[0]
y
y.apply(lambda x: x.item().total_seconds())

also,

y.apply(lambda x: x)

should return y, but it returns a Series of ints, open a GH issue?

[cpcloud]

only python's native timedelta objects would satisfy this code. should i remove this example? or i could wrap it in timedelta calls

[jreback]

no...that's ok...its meant for numpy 1.6.2 (if you look at the docs they build ok), I had to put a note that the syntax is diff in 1.7; I didn't want to put a numpy version detector, seemed like too much magic

[cpcloud]

hm maybe could be a sphinx argument tho to the ipython directive like :numpy==1.6: similar to pip. easy enough to parse maybe i will do that

[jreback]

sure.....(and the 1.7 replacements for those 2 statements are in the note....)

this is because numpy changed how timedelta64[ns] works in 1.7 vs. 1.6.2

[cpcloud]

cool sort of flew through and didn't see the note. i was just using grin and vim so i just went straight to prob. saw the td64 change in the np docs.

[jreback]

I'll have to say though, that I do sometimes catch bugs because the docs throw build errors! (and then I make that a test)....so its good to build them occasionally

[jreback]

FYI...all of your html stuff building nicely now on public docs...I threw in some links back to the docs (will buidl tomorrow as I just pushed it)....I think all of your most recent changes are in there though

[cpcloud]

sweet thanks

[cpcloud]

alright i've squashed as many of these warnings as i can. i've converted all of the uses of reset_printoptions to reset_option and the same for set_printoptions

[jreback]

were they in the main docs? or in whatsnew?

[cpcloud]

main docs

[cpcloud]

and in a few tests

[cpcloud]

with the exception of a warning from a call to store.unique in v0.10.1.txt (now deprecated). i added a warning suppression. i can remove that if u want

[jreback]

no that's good
no reason to have the warning in the doc build

[cpcloud]

pr is up